16 files changed, 214 insertions, 130 deletions

diff --git a/drivers/gpu/drm/i915/gvt/cfg_space.c b/drivers/gpu/drm/i915/gvt/cfg_space.c
index 97bfc00d2a82..c62346fdc05d 100644
--- a/drivers/gpu/drm/i915/gvt/cfg_space.c
+++ b/drivers/gpu/drm/i915/gvt/cfg_space.c
@@ -119,16 +119,6 @@ static int map_aperture(struct intel_vgpu *vgpu, bool map)
 	if (map == vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked)
 		return 0;
 
-	if (map) {
-		vgpu->gm.aperture_va = memremap(aperture_pa, aperture_sz,
-						MEMREMAP_WC);
-		if (!vgpu->gm.aperture_va)
-			return -ENOMEM;
-	} else {
-		memunmap(vgpu->gm.aperture_va);
-		vgpu->gm.aperture_va = NULL;
-	}
-
 	val = vgpu_cfg_space(vgpu)[PCI_BASE_ADDRESS_2];
 	if (val & PCI_BASE_ADDRESS_MEM_TYPE_64)
 		val = *(u64 *)(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_2);
@@ -141,11 +131,8 @@ static int map_aperture(struct intel_vgpu *vgpu, bool map)
 						  aperture_pa >> PAGE_SHIFT,
 						  aperture_sz >> PAGE_SHIFT,
 						  map);
-	if (ret) {
-		memunmap(vgpu->gm.aperture_va);
-		vgpu->gm.aperture_va = NULL;
+	if (ret)
 		return ret;
-	}
 
 	vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked = map;
 	return 0;
diff --git a/drivers/gpu/drm/i915/gvt/dmabuf.c b/drivers/gpu/drm/i915/gvt/dmabuf.c
index 2ab584f97dfb..2fb7b34ef561 100644
--- a/drivers/gpu/drm/i915/gvt/dmabuf.c
+++ b/drivers/gpu/drm/i915/gvt/dmabuf.c
@@ -472,7 +472,6 @@ int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id)
 		ret = PTR_ERR(dmabuf);
 		goto out_free_gem;
 	}
-	obj->base.dma_buf = dmabuf;
 
 	i915_gem_object_put(obj);
 
diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c
index 769c1c24ae75..70494e394d2c 100644
--- a/drivers/gpu/drm/i915/gvt/execlist.c
+++ b/drivers/gpu/drm/i915/gvt/execlist.c
@@ -521,24 +521,23 @@ static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id)
 
 	ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
 			_EL_OFFSET_STATUS_PTR);
-
 	ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);
 	ctx_status_ptr.read_ptr = 0;
 	ctx_status_ptr.write_ptr = 0x7;
 	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
 }
 
-static void clean_execlist(struct intel_vgpu *vgpu)
+static void clean_execlist(struct intel_vgpu *vgpu, unsigned long engine_mask)
 {
-	enum intel_engine_id i;
+	unsigned int tmp;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 	struct intel_engine_cs *engine;
+	struct intel_vgpu_submission *s = &vgpu->submission;
 
-	for_each_engine(engine, vgpu->gvt->dev_priv, i) {
-		struct intel_vgpu_submission *s = &vgpu->submission;
-
-		kfree(s->ring_scan_buffer[i]);
-		s->ring_scan_buffer[i] = NULL;
-		s->ring_scan_buffer_size[i] = 0;
+	for_each_engine_masked(engine, dev_priv, engine_mask, tmp) {
+		kfree(s->ring_scan_buffer[engine->id]);
+		s->ring_scan_buffer[engine->id] = NULL;
+		s->ring_scan_buffer_size[engine->id] = 0;
 	}
 }
 
@@ -553,9 +552,10 @@ static void reset_execlist(struct intel_vgpu *vgpu,
 		init_vgpu_execlist(vgpu, engine->id);
 }
 
-static int init_execlist(struct intel_vgpu *vgpu)
+static int init_execlist(struct intel_vgpu *vgpu,
+			 unsigned long engine_mask)
 {
-	reset_execlist(vgpu, ALL_ENGINES);
+	reset_execlist(vgpu, engine_mask);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index a529d2bd393c..8d5317d0122d 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -997,9 +997,11 @@ static inline void ppgtt_generate_shadow_entry(struct intel_gvt_gtt_entry *se,
 static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 {
 	struct intel_vgpu *vgpu = spt->vgpu;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
 	struct intel_vgpu_ppgtt_spt *s;
 	struct intel_gvt_gtt_entry se, ge;
-	unsigned long i;
+	unsigned long gfn, i;
 	int ret;
 
 	trace_spt_change(spt->vgpu->id, "born", spt,
@@ -1007,9 +1009,10 @@ static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 
 	if (gtt_type_is_pte_pt(spt->shadow_page.type)) {
 		for_each_present_guest_entry(spt, &ge, i) {
-			ret = gtt_entry_p2m(vgpu, &ge, &se);
-			if (ret)
-				goto fail;
+			gfn = ops->get_pfn(&ge);
+			if (!intel_gvt_hypervisor_is_valid_gfn(vgpu, gfn) ||
+				gtt_entry_p2m(vgpu, &ge, &se))
+				ops->set_pfn(&se, gvt->gtt.scratch_mfn);
 			ppgtt_set_shadow_entry(spt, &se, i);
 		}
 		return 0;
@@ -1906,7 +1909,7 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 	struct intel_vgpu_mm *ggtt_mm = vgpu->gtt.ggtt_mm;
 	struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
 	unsigned long g_gtt_index = off >> info->gtt_entry_size_shift;
-	unsigned long gma;
+	unsigned long gma, gfn;
 	struct intel_gvt_gtt_entry e, m;
 	int ret;
 
@@ -1925,6 +1928,16 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 			bytes);
 
 	if (ops->test_present(&e)) {
+		gfn = ops->get_pfn(&e);
+
+		/* one PTE update may be issued in multiple writes and the
+		 * first write may not construct a valid gfn
+		 */
+		if (!intel_gvt_hypervisor_is_valid_gfn(vgpu, gfn)) {
+			ops->set_pfn(&m, gvt->gtt.scratch_mfn);
+			goto out;
+		}
+
 		ret = gtt_entry_p2m(vgpu, &e, &m);
 		if (ret) {
 			gvt_vgpu_err("fail to translate guest gtt entry\n");
@@ -1939,6 +1952,7 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 		ops->set_pfn(&m, gvt->gtt.scratch_mfn);
 	}
 
+out:
 	ggtt_set_shadow_entry(ggtt_mm, &m, g_gtt_index);
 	gtt_invalidate(gvt->dev_priv);
 	ggtt_set_guest_entry(ggtt_mm, &e, g_gtt_index);
diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
index 7dc7a80213a8..c6197d990818 100644
--- a/drivers/gpu/drm/i915/gvt/gvt.h
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -82,7 +82,6 @@ struct intel_gvt_device_info {
 struct intel_vgpu_gm {
 	u64 aperture_sz;
 	u64 hidden_sz;
-	void *aperture_va;
 	struct drm_mm_node low_gm_node;
 	struct drm_mm_node high_gm_node;
 };
@@ -127,7 +126,6 @@ struct intel_vgpu_irq {
 struct intel_vgpu_opregion {
 	bool mapped;
 	void *va;
-	void *va_gopregion;
 	u32 gfn[INTEL_GVT_OPREGION_PAGES];
 };
 
@@ -152,8 +150,8 @@ enum {
 
 struct intel_vgpu_submission_ops {
 	const char *name;
-	int (*init)(struct intel_vgpu *vgpu);
-	void (*clean)(struct intel_vgpu *vgpu);
+	int (*init)(struct intel_vgpu *vgpu, unsigned long engine_mask);
+	void (*clean)(struct intel_vgpu *vgpu, unsigned long engine_mask);
 	void (*reset)(struct intel_vgpu *vgpu, unsigned long engine_mask);
 };
 
diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c
index 92d6468daeee..9be639aa3b55 100644
--- a/drivers/gpu/drm/i915/gvt/handlers.c
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -1494,7 +1494,6 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
-	struct intel_vgpu_submission *s = &vgpu->submission;
 	u32 data = *(u32 *)p_data;
 	int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset);
 	bool enable_execlist;
@@ -1523,11 +1522,9 @@ static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		if (!enable_execlist)
 			return 0;
 
-		if (s->active)
-			return 0;
-
 		ret = intel_vgpu_select_submission_ops(vgpu,
-				INTEL_VGPU_EXECLIST_SUBMISSION);
+			       ENGINE_MASK(ring_id),
+			       INTEL_VGPU_EXECLIST_SUBMISSION);
 		if (ret)
 			return ret;
 
@@ -2843,6 +2840,9 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(_MMIO(_PLANE_KEYVAL_1(PIPE_A)), D_SKL_PLUS);
 	MMIO_D(_MMIO(_PLANE_KEYVAL_1(PIPE_B)), D_SKL_PLUS);
 	MMIO_D(_MMIO(_PLANE_KEYVAL_1(PIPE_C)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYMAX_1(PIPE_A)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYMAX_1(PIPE_B)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYMAX_1(PIPE_C)), D_SKL_PLUS);
 	MMIO_D(_MMIO(_PLANE_KEYMSK_1(PIPE_A)), D_SKL_PLUS);
 	MMIO_D(_MMIO(_PLANE_KEYMSK_1(PIPE_B)), D_SKL_PLUS);
 	MMIO_D(_MMIO(_PLANE_KEYMSK_1(PIPE_C)), D_SKL_PLUS);
diff --git a/drivers/gpu/drm/i915/gvt/hypercall.h b/drivers/gpu/drm/i915/gvt/hypercall.h
index a1bd82feb827..f8e77e166246 100644
--- a/drivers/gpu/drm/i915/gvt/hypercall.h
+++ b/drivers/gpu/drm/i915/gvt/hypercall.h
@@ -58,6 +58,7 @@ struct intel_gvt_mpt {
 	int (*set_opregion)(void *vgpu);
 	int (*get_vfio_device)(void *vgpu);
 	void (*put_vfio_device)(void *vgpu);
+	bool (*is_valid_gfn)(unsigned long handle, unsigned long gfn);
 };
 
 extern struct intel_gvt_mpt xengt_mpt;
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index 554d1db1f3c8..909499b73d03 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -651,6 +651,39 @@ static int intel_vgpu_bar_rw(struct intel_vgpu *vgpu, int bar, uint64_t off,
 	return ret;
 }
 
+static inline bool intel_vgpu_in_aperture(struct intel_vgpu *vgpu, uint64_t off)
+{
+	return off >= vgpu_aperture_offset(vgpu) &&
+	       off < vgpu_aperture_offset(vgpu) + vgpu_aperture_sz(vgpu);
+}
+
+static int intel_vgpu_aperture_rw(struct intel_vgpu *vgpu, uint64_t off,
+		void *buf, unsigned long count, bool is_write)
+{
+	void *aperture_va;
+
+	if (!intel_vgpu_in_aperture(vgpu, off) ||
+	    !intel_vgpu_in_aperture(vgpu, off + count)) {
+		gvt_vgpu_err("Invalid aperture offset %llu\n", off);
+		return -EINVAL;
+	}
+
+	aperture_va = io_mapping_map_wc(&vgpu->gvt->dev_priv->ggtt.iomap,
+					ALIGN_DOWN(off, PAGE_SIZE),
+					count + offset_in_page(off));
+	if (!aperture_va)
+		return -EIO;
+
+	if (is_write)
+		memcpy(aperture_va + offset_in_page(off), buf, count);
+	else
+		memcpy(buf, aperture_va + offset_in_page(off), count);
+
+	io_mapping_unmap(aperture_va);
+
+	return 0;
+}
+
 static ssize_t intel_vgpu_rw(struct mdev_device *mdev, char *buf,
 			size_t count, loff_t *ppos, bool is_write)
 {
@@ -679,8 +712,7 @@ static ssize_t intel_vgpu_rw(struct mdev_device *mdev, char *buf,
 					buf, count, is_write);
 		break;
 	case VFIO_PCI_BAR2_REGION_INDEX:
-		ret = intel_vgpu_bar_rw(vgpu, PCI_BASE_ADDRESS_2, pos,
-					buf, count, is_write);
+		ret = intel_vgpu_aperture_rw(vgpu, pos, buf, count, is_write);
 		break;
 	case VFIO_PCI_BAR1_REGION_INDEX:
 	case VFIO_PCI_BAR3_REGION_INDEX:
@@ -1575,6 +1607,21 @@ static unsigned long kvmgt_virt_to_pfn(void *addr)
 	return PFN_DOWN(__pa(addr));
 }
 
+static bool kvmgt_is_valid_gfn(unsigned long handle, unsigned long gfn)
+{
+	struct kvmgt_guest_info *info;
+	struct kvm *kvm;
+
+	if (!handle_valid(handle))
+		return false;
+
+	info = (struct kvmgt_guest_info *)handle;
+	kvm = info->kvm;
+
+	return kvm_is_visible_gfn(kvm, gfn);
+
+}
+
 struct intel_gvt_mpt kvmgt_mpt = {
 	.host_init = kvmgt_host_init,
 	.host_exit = kvmgt_host_exit,
@@ -1590,6 +1637,7 @@ struct intel_gvt_mpt kvmgt_mpt = {
 	.set_opregion = kvmgt_set_opregion,
 	.get_vfio_device = kvmgt_get_vfio_device,
 	.put_vfio_device = kvmgt_put_vfio_device,
+	.is_valid_gfn = kvmgt_is_valid_gfn,
 };
 EXPORT_SYMBOL_GPL(kvmgt_mpt);
 
diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c
index 562b5ad857a4..5c869e3fdf3b 100644
--- a/drivers/gpu/drm/i915/gvt/mmio.c
+++ b/drivers/gpu/drm/i915/gvt/mmio.c
@@ -56,38 +56,6 @@ int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa)
 	(reg >= gvt->device_info.gtt_start_offset \
 	 && reg < gvt->device_info.gtt_start_offset + gvt_ggtt_sz(gvt))
 
-static bool vgpu_gpa_is_aperture(struct intel_vgpu *vgpu, uint64_t gpa)
-{
-	u64 aperture_gpa = intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_2);
-	u64 aperture_sz = vgpu_aperture_sz(vgpu);
-
-	return gpa >= aperture_gpa && gpa < aperture_gpa + aperture_sz;
-}
-
-static int vgpu_aperture_rw(struct intel_vgpu *vgpu, uint64_t gpa,
-			    void *pdata, unsigned int size, bool is_read)
-{
-	u64 aperture_gpa = intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_2);
-	u64 offset = gpa - aperture_gpa;
-
-	if (!vgpu_gpa_is_aperture(vgpu, gpa + size - 1)) {
-		gvt_vgpu_err("Aperture rw out of range, offset %llx, size %d\n",
-			     offset, size);
-		return -EINVAL;
-	}
-
-	if (!vgpu->gm.aperture_va) {
-		gvt_vgpu_err("BAR is not enabled\n");
-		return -ENXIO;
-	}
-
-	if (is_read)
-		memcpy(pdata, vgpu->gm.aperture_va + offset, size);
-	else
-		memcpy(vgpu->gm.aperture_va + offset, pdata, size);
-	return 0;
-}
-
 static void failsafe_emulate_mmio_rw(struct intel_vgpu *vgpu, uint64_t pa,
 		void *p_data, unsigned int bytes, bool read)
 {
@@ -144,11 +112,6 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 	}
 	mutex_lock(&gvt->lock);
 
-	if (vgpu_gpa_is_aperture(vgpu, pa)) {
-		ret = vgpu_aperture_rw(vgpu, pa, p_data, bytes, true);
-		goto out;
-	}
-
 	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
 
 	if (WARN_ON(bytes > 8))
@@ -222,11 +185,6 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 
 	mutex_lock(&gvt->lock);
 
-	if (vgpu_gpa_is_aperture(vgpu, pa)) {
-		ret = vgpu_aperture_rw(vgpu, pa, p_data, bytes, false);
-		goto out;
-	}
-
 	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
 
 	if (WARN_ON(bytes > 8))
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
index 74834395dd89..73ad6e90e49d 100644
--- a/drivers/gpu/drm/i915/gvt/mmio_context.c
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -80,7 +80,7 @@ static struct engine_mmio gen8_engine_mmio_list[] __cacheline_aligned = {
 	{BCS, RING_INSTPM(BLT_RING_BASE), 0xffff, false}, /* 0x220c0 */
 	{BCS, RING_HWSTAM(BLT_RING_BASE), 0x0, false}, /* 0x22098 */
 	{BCS, RING_EXCC(BLT_RING_BASE), 0x0, false}, /* 0x22028 */
-	{ /* Terminated */ }
+	{RCS, INVALID_MMIO_REG, 0, false } /* Terminated */
 };
 
 static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = {
@@ -146,7 +146,7 @@ static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = {
 	{RCS, GEN8_GARBCNTL, 0x0, false}, /* 0xb004 */
 	{RCS, GEN7_FF_THREAD_MODE, 0x0, false}, /* 0x20a0 */
 	{RCS, FF_SLICE_CS_CHICKEN2, 0xffff, false}, /* 0x20e4 */
-	{ /* Terminated */ }
+	{RCS, INVALID_MMIO_REG, 0, false } /* Terminated */
 };
 
 static struct {
@@ -167,7 +167,7 @@ static void load_render_mocs(struct drm_i915_private *dev_priv)
 	};
 	int ring_id, i;
 
-	for (ring_id = 0; ring_id < I915_NUM_ENGINES; ring_id++) {
+	for (ring_id = 0; ring_id < ARRAY_SIZE(regs); ring_id++) {
 		offset.reg = regs[ring_id];
 		for (i = 0; i < 64; i++) {
 			gen9_render_mocs.control_table[ring_id][i] =
@@ -310,8 +310,8 @@ static void switch_mmio(struct intel_vgpu *pre,
 	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
 		switch_mocs(pre, next, ring_id);
 
-	mmio = dev_priv->gvt->engine_mmio_list;
-	while (i915_mmio_reg_offset((mmio++)->reg)) {
+	for (mmio = dev_priv->gvt->engine_mmio_list;
+	     i915_mmio_reg_valid(mmio->reg); mmio++) {
 		if (mmio->ring_id != ring_id)
 			continue;
 		// save
diff --git a/drivers/gpu/drm/i915/gvt/mpt.h b/drivers/gpu/drm/i915/gvt/mpt.h
index ca8005a6d5fa..81aff4eacbfe 100644
--- a/drivers/gpu/drm/i915/gvt/mpt.h
+++ b/drivers/gpu/drm/i915/gvt/mpt.h
@@ -339,4 +339,21 @@ static inline void intel_gvt_hypervisor_put_vfio_device(struct intel_vgpu *vgpu)
 	intel_gvt_host.mpt->put_vfio_device(vgpu);
 }
 
+/**
+ * intel_gvt_hypervisor_is_valid_gfn - check if a visible gfn
+ * @vgpu: a vGPU
+ * @gfn: guest PFN
+ *
+ * Returns:
+ * true on valid gfn, false on not.
+ */
+static inline bool intel_gvt_hypervisor_is_valid_gfn(
+		struct intel_vgpu *vgpu, unsigned long gfn)
+{
+	if (!intel_gvt_host.mpt->is_valid_gfn)
+		return true;
+
+	return intel_gvt_host.mpt->is_valid_gfn(vgpu->handle, gfn);
+}
+
 #endif /* _GVT_MPT_H_ */
diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c
index 8420d1fc3ddb..fa75a2eead90 100644
--- a/drivers/gpu/drm/i915/gvt/opregion.c
+++ b/drivers/gpu/drm/i915/gvt/opregion.c
@@ -299,21 +299,13 @@ int intel_vgpu_opregion_base_write_handler(struct intel_vgpu *vgpu, u32 gpa)
 {
 
 	int i, ret = 0;
-	unsigned long pfn;
 
 	gvt_dbg_core("emulate opregion from kernel\n");
 
 	switch (intel_gvt_host.hypervisor_type) {
 	case INTEL_GVT_HYPERVISOR_KVM:
-		pfn = intel_gvt_hypervisor_gfn_to_mfn(vgpu, gpa >> PAGE_SHIFT);
-		vgpu_opregion(vgpu)->va_gopregion = memremap(pfn << PAGE_SHIFT,
-						INTEL_GVT_OPREGION_SIZE,
-						MEMREMAP_WB);
-		if (!vgpu_opregion(vgpu)->va_gopregion) {
-			gvt_vgpu_err("failed to map guest opregion\n");
-			ret = -EFAULT;
-		}
-		vgpu_opregion(vgpu)->mapped = true;
+		for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
+			vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
 		break;
 	case INTEL_GVT_HYPERVISOR_XEN:
 		/**
@@ -352,10 +344,7 @@ void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu)
 		if (vgpu_opregion(vgpu)->mapped)
 			map_vgpu_opregion(vgpu, false);
 	} else if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_KVM) {
-		if (vgpu_opregion(vgpu)->mapped) {
-			memunmap(vgpu_opregion(vgpu)->va_gopregion);
-			vgpu_opregion(vgpu)->va_gopregion = NULL;
-		}
+		/* Guest opregion is released by VFIO */
 	}
 	free_pages((unsigned long)vgpu_opregion(vgpu)->va,
 		   get_order(INTEL_GVT_OPREGION_SIZE));
@@ -480,19 +469,40 @@ static bool querying_capabilities(u32 scic)
  */
 int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci)
 {
-	u32 *scic, *parm;
+	u32 scic, parm;
 	u32 func, subfunc;
+	u64 scic_pa = 0, parm_pa = 0;
+	int ret;
 
 	switch (intel_gvt_host.hypervisor_type) {
 	case INTEL_GVT_HYPERVISOR_XEN:
-		scic = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_SCIC;
-		parm = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_PARM;
+		scic = *((u32 *)vgpu_opregion(vgpu)->va +
+					INTEL_GVT_OPREGION_SCIC);
+		parm = *((u32 *)vgpu_opregion(vgpu)->va +
+					INTEL_GVT_OPREGION_PARM);
 		break;
 	case INTEL_GVT_HYPERVISOR_KVM:
-		scic = vgpu_opregion(vgpu)->va_gopregion +
-						INTEL_GVT_OPREGION_SCIC;
-		parm = vgpu_opregion(vgpu)->va_gopregion +
-						INTEL_GVT_OPREGION_PARM;
+		scic_pa = (vgpu_opregion(vgpu)->gfn[0] << PAGE_SHIFT) +
+					INTEL_GVT_OPREGION_SCIC;
+		parm_pa = (vgpu_opregion(vgpu)->gfn[0] << PAGE_SHIFT) +
+					INTEL_GVT_OPREGION_PARM;
+
+		ret = intel_gvt_hypervisor_read_gpa(vgpu, scic_pa,
+						    &scic, sizeof(scic));
+		if (ret) {
+			gvt_vgpu_err("guest opregion read error %d, gpa 0x%llx, len %lu\n",
+				ret, scic_pa, sizeof(scic));
+			return ret;
+		}
+
+		ret = intel_gvt_hypervisor_read_gpa(vgpu, parm_pa,
+						    &parm, sizeof(parm));
+		if (ret) {
+			gvt_vgpu_err("guest opregion read error %d, gpa 0x%llx, len %lu\n",
+				ret, scic_pa, sizeof(scic));
+			return ret;
+		}
+
 		break;
 	default:
 		gvt_vgpu_err("not supported hypervisor\n");
@@ -510,9 +520,9 @@ int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci)
 		return 0;
 	}
 
-	func = GVT_OPREGION_FUNC(*scic);
-	subfunc = GVT_OPREGION_SUBFUNC(*scic);
-	if (!querying_capabilities(*scic)) {
+	func = GVT_OPREGION_FUNC(scic);
+	subfunc = GVT_OPREGION_SUBFUNC(scic);
+	if (!querying_capabilities(scic)) {
 		gvt_vgpu_err("requesting runtime service: func \"%s\","
 				" subfunc \"%s\"\n",
 				opregion_func_name(func),
@@ -521,11 +531,43 @@ int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci)
 		 * emulate exit status of function call, '0' means
 		 * "failure, generic, unsupported or unknown cause"
 		 */
-		*scic &= ~OPREGION_SCIC_EXIT_MASK;
-		return 0;
+		scic &= ~OPREGION_SCIC_EXIT_MASK;
+		goto out;
+	}
+
+	scic = 0;
+	parm = 0;
+
+out:
+	switch (intel_gvt_host.hypervisor_type) {
+	case INTEL_GVT_HYPERVISOR_XEN:
+		*((u32 *)vgpu_opregion(vgpu)->va +
+					INTEL_GVT_OPREGION_SCIC) = scic;
+		*((u32 *)vgpu_opregion(vgpu)->va +
+					INTEL_GVT_OPREGION_PARM) = parm;
+		break;
+	case INTEL_GVT_HYPERVISOR_KVM:
+		ret = intel_gvt_hypervisor_write_gpa(vgpu, scic_pa,
+						    &scic, sizeof(scic));
+		if (ret) {
+			gvt_vgpu_err("guest opregion write error %d, gpa 0x%llx, len %lu\n",
+				ret, scic_pa, sizeof(scic));
+			return ret;
+		}
+
+		ret = intel_gvt_hypervisor_write_gpa(vgpu, parm_pa,
+						    &parm, sizeof(parm));
+		if (ret) {
+			gvt_vgpu_err("guest opregion write error %d, gpa 0x%llx, len %lu\n",
+				ret, scic_pa, sizeof(scic));
+			return ret;
+		}
+
+		break;
+	default:
+		gvt_vgpu_err("not supported hypervisor\n");
+		return -EINVAL;
 	}
 
-	*scic = 0;
-	*parm = 0;
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/gvt/sched_policy.c b/drivers/gpu/drm/i915/gvt/sched_policy.c
index eea1a2f92099..cc1ce361cd76 100644
--- a/drivers/gpu/drm/i915/gvt/sched_policy.c
+++ b/drivers/gpu/drm/i915/gvt/sched_policy.c
@@ -50,6 +50,7 @@ static bool vgpu_has_pending_workload(struct intel_vgpu *vgpu)
 struct vgpu_sched_data {
 	struct list_head lru_list;
 	struct intel_vgpu *vgpu;
+	bool active;
 
 	ktime_t sched_in_time;
 	ktime_t sched_out_time;
@@ -308,8 +309,15 @@ static int tbs_sched_init_vgpu(struct intel_vgpu *vgpu)
 
 static void tbs_sched_clean_vgpu(struct intel_vgpu *vgpu)
 {
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct gvt_sched_data *sched_data = gvt->scheduler.sched_data;
+
 	kfree(vgpu->sched_data);
 	vgpu->sched_data = NULL;
+
+	/* this vgpu id has been removed */
+	if (idr_is_empty(&gvt->vgpu_idr))
+		hrtimer_cancel(&sched_data->timer);
 }
 
 static void tbs_sched_start_schedule(struct intel_vgpu *vgpu)
@@ -325,6 +333,7 @@ static void tbs_sched_start_schedule(struct intel_vgpu *vgpu)
 	if (!hrtimer_active(&sched_data->timer))
 		hrtimer_start(&sched_data->timer, ktime_add_ns(ktime_get(),
 			sched_data->period), HRTIMER_MODE_ABS);
+	vgpu_data->active = true;
 }
 
 static void tbs_sched_stop_schedule(struct intel_vgpu *vgpu)
@@ -332,6 +341,7 @@ static void tbs_sched_stop_schedule(struct intel_vgpu *vgpu)
 	struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
 
 	list_del_init(&vgpu_data->lru_list);
+	vgpu_data->active = false;
 }
 
 static struct intel_gvt_sched_policy_ops tbs_schedule_ops = {
@@ -367,9 +377,12 @@ void intel_vgpu_clean_sched_policy(struct intel_vgpu *vgpu)
 
 void intel_vgpu_start_schedule(struct intel_vgpu *vgpu)
 {
-	gvt_dbg_core("vgpu%d: start schedule\n", vgpu->id);
+	struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
 
-	vgpu->gvt->scheduler.sched_ops->start_schedule(vgpu);
+	if (!vgpu_data->active) {
+		gvt_dbg_core("vgpu%d: start schedule\n", vgpu->id);
+		vgpu->gvt->scheduler.sched_ops->start_schedule(vgpu);
+	}
 }
 
 void intel_gvt_kick_schedule(struct intel_gvt *gvt)
@@ -382,6 +395,10 @@ void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu)
 	struct intel_gvt_workload_scheduler *scheduler =
 		&vgpu->gvt->scheduler;
 	int ring_id;
+	struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
+
+	if (!vgpu_data->active)
+		return;
 
 	gvt_dbg_core("vgpu%d: stop schedule\n", vgpu->id);
 
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 0056638b0c16..b55b3580ca1d 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -991,7 +991,7 @@ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu)
 {
 	struct intel_vgpu_submission *s = &vgpu->submission;
 
-	intel_vgpu_select_submission_ops(vgpu, 0);
+	intel_vgpu_select_submission_ops(vgpu, ALL_ENGINES, 0);
 	i915_gem_context_put(s->shadow_ctx);
 	kmem_cache_destroy(s->workloads);
 }
@@ -1079,6 +1079,7 @@ out_shadow_ctx:
  *
  */
 int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+				     unsigned long engine_mask,
 				     unsigned int interface)
 {
 	struct intel_vgpu_submission *s = &vgpu->submission;
@@ -1091,21 +1092,21 @@ int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
 	if (WARN_ON(interface >= ARRAY_SIZE(ops)))
 		return -EINVAL;
 
-	if (s->active) {
-		s->ops->clean(vgpu);
-		s->active = false;
-		gvt_dbg_core("vgpu%d: de-select ops [ %s ] \n",
-				vgpu->id, s->ops->name);
-	}
+	if (WARN_ON(interface == 0 && engine_mask != ALL_ENGINES))
+		return -EINVAL;
+
+	if (s->active)
+		s->ops->clean(vgpu, engine_mask);
 
 	if (interface == 0) {
 		s->ops = NULL;
 		s->virtual_submission_interface = 0;
-		gvt_dbg_core("vgpu%d: no submission ops\n", vgpu->id);
+		s->active = false;
+		gvt_dbg_core("vgpu%d: remove submission ops\n", vgpu->id);
 		return 0;
 	}
 
-	ret = ops[interface]->init(vgpu);
+	ret = ops[interface]->init(vgpu, engine_mask);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h
index 3de77dfa7c59..ff175a98b19e 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.h
+++ b/drivers/gpu/drm/i915/gvt/scheduler.h
@@ -141,6 +141,7 @@ void intel_vgpu_reset_submission(struct intel_vgpu *vgpu,
 void intel_vgpu_clean_submission(struct intel_vgpu *vgpu);
 
 int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+				     unsigned long engine_mask,
 				     unsigned int interface);
 
 extern const struct intel_vgpu_submission_ops
diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c
index 4688619f6a1c..b87b19d8443c 100644
--- a/drivers/gpu/drm/i915/gvt/vgpu.c
+++ b/drivers/gpu/drm/i915/gvt/vgpu.c
@@ -258,6 +258,8 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 
 	intel_gvt_debugfs_remove_vgpu(vgpu);
 	idr_remove(&gvt->vgpu_idr, vgpu->id);
+	if (idr_is_empty(&gvt->vgpu_idr))
+		intel_gvt_clean_irq(gvt);
 	intel_vgpu_clean_sched_policy(vgpu);
 	intel_vgpu_clean_submission(vgpu);
 	intel_vgpu_clean_display(vgpu);
@@ -518,8 +520,7 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
 	intel_vgpu_reset_submission(vgpu, resetting_eng);
 	/* full GPU reset or device model level reset */
 	if (engine_mask == ALL_ENGINES || dmlr) {
-		intel_vgpu_select_submission_ops(vgpu, 0);
-
+		intel_vgpu_select_submission_ops(vgpu, ALL_ENGINES, 0);
 		/*fence will not be reset during virtual reset */
 		if (dmlr) {
 			intel_vgpu_reset_gtt(vgpu);